Canonicalization: SEO Mythbusting
RACHEL COSTELLO: I think
that a lot of people see canonicalization as kind
of topical grouping, which kind of isn't right at all. They need to, the pages need
to be either identical or near. MARTIN SPLITT: Near
identical, exactly. RACHEL COSTELLO: Yeah. MARTIN SPLITT: Yeah That's
what it boils down to. Canonicalization is about
duplication management. So basically, you want
to remove duplication, so that we don't have to
crawl things multiple times and we don't have to render and
index things multiple times. And we also do not
serve them all the time, like the same things basically
in three different URLs. That's not good search
results really, right? [MUSIC PLAYING] Hello, everybody, and
welcome to another episode of "SEO Mythbusting." With me today is
Rachel Costello. You are a DeepCrawl Technical
SEO and Content Manager.
So what is it that
you're doing every day? RACHEL COSTELLO: So
I basically, well, I used to be a
technical SEO myself, and now I've moved into more
of the content production side of things. So writing white
papers, articles, to educate the wider dev
community, digital marketing community, about technical
SEO and the impact it has. MARTIN SPLITT: Awesome. That's really interesting. So you're seeing a bunch of
misconceptions and confusions and stuff. And we picked an interesting
topic, didn't we? RACHEL COSTELLO: We did. MARTIN SPLITT: What's
the topic that you want to talk about today? RACHEL COSTELLO: It's all
about canonicalization. MARTIN SPLITT: Ooh. All right. So what are the top
myths and misconceptions that the community
is dealing with? RACHEL COSTELLO: So I think
the first thing is that people think it is a directive.
You set a canonical tag. It's going to be accepted. Another one, yeah exactly. Another one is that they kind
of use it like a redirect. So if you have a product
page that goes out of stock, you add a canonical to
that category page, which doesn't really work that way. Because I've heard that the
content needs to ideally be identical, if
not very similar. MARTIN SPLITT: Yeah. RACHEL COSTELLO: So lots
of things like that. MARTIN SPLITT: Oh, interesting. All right, let's
start with the idea that it is a directive,
because it's not. RACHEL COSTELLO: No. MARTIN SPLITT: No. It is a signal for us, right? So when we talk about
canonicalization, we're talking about detecting
content or the same content or a very similar content that
exists on different addresses and the different URLs, right? So we can do many different
things to basically identify these things, right? We can just crawl multiple
pages and see like, oh, this is actually the same content. We can also probably
see if the same links and like the same kind
of context is used.
But also, we can use the
canonical tag, right? It's a signal. We're using many
different signals to figure out if something
is the same content or not. And canonicalization
with a canonical tag is just one of them. So putting a canonical tag on
pages that are not the same is not going to work. Putting a canonical tag
on each of the pages that are exactly the same is
also not going to work. It is a signal. It helps us identify what
we want to canonicalize, but it doesn't say
you have to use this. That's a big one, I think. And you're right. And you should not use it
as a redirection either.
It's not a redirect. RACHEL COSTELLO: I think people
just want to group link equity wherever they can,
and it's maybe a bit of a desperate act to
try and keep all of their link equity in one place. MARTIN SPLITT: It is, it is. Again, like
canonicalization makes sense if you cross post
the same content on different, I don't
know, platforms, or different channels in
slightly different locations for whatever reason
you're doing that. That's where
canonicalization comes in. But if you are having something
that goes out of stock, you should either redirect
it to something similar that makes sense for the user at that
point, or you can just tell us, this is a 404 for the
moment and might come back. But do not just
think that you can, no, it's not the same
as a redirection.
Also, you're wasting
crawl budget that way. Because we are just
not understanding like, oh, so you're saying this
is the same as the other page, but it clearly is not. So we're just going to
continue doing this. But if you have two
pages that are identical and you're not canonicalization
or you're not canonicalizing them the way that
it makes sense, then we kind of have to
look into both as well. And sometimes we get these
like flipping canonicals. Yeah. What are the typical
problems that you're seeing that people are having
besides these misconceptions? Like, what are people doing with
them you think makes no sense? RACHEL COSTELLO: So I think
people are just not quite sure. We've been trying
to piece together what these different signals are
that play into [? effects.
?] You've got redirect, site maps,
backlinks and things like that. I think people are
trying to weigh up how many of these
signals they should add. Maybe they're kind of doing
it like a mass equation. Like if I do these
two things, then this will mean that Google picks
my canonical tag that I want. But it would just
be interesting.
I'm always interested
to know more about how the signals are
weighted, which ones are more preferential to others. Because sometimes
I see that maybe, this is just my theory,
that maybe Google puts more weighting
to signals that are more likely to
have been implemented by human rather than maybe
an auto-generated setting. I don't know if that has any. MARTIN SPLITT: Well,
duplication and deduplication is actually done without
much human interaction. So this is all
automated signals. But we do like content
fingerprinting. We look at things like,
what is the gist of it really, what are the,
what's the information here, how does this relate to
the site structure, what does it say in the site map. So we're looking at a
bunch of different factors, but they're mostly
technical factors. RACHEL COSTELLO: OK. MARTIN SPLITT: Yeah. And we are basically scoring
them on an ongoing basis. So it's not that we're
like determining at once and then just stick to it.
We are always looking
at the fresh content that we got from crawl, and
then have a look at, does this change, is it changed,
is it now very close to what it has been before. Now maybe something that
has been in duplication is no longer a
duplicate, because it has changed its contents. So that is absolutely
possible, right? But sometimes, especially
when pretty much everything is showing up in the
same URL structure and it's maybe like
different language versions of the same thing,
but it is the same content, then we might end up with a
scoring that is very similar. So we have both versions. And let's say like
one 0.49 and one is 0.51 of what we think is
a duplication of the other, then it's really
hard to pick which one will be the canonical. And that can change, right? A change in, I don't
know, how we crawl things or how the crawler
has fetch data and how it has been
touching the other pages beforehand might influence us
to have like a tiny little bit of a jump in these two numbers.
And then the other
one is the canonical. So make sure that you're trying
to give us as clear a signal as possible and not confuse the
algorithms that are working with figuring out which one
is the duplication of which other thing. Because if we are having
two equal pieces of content, then how do we know
which one we should pick? RACHEL COSTELLO: Exactly. And you don't want Google
to be in that position where they feel like
they have to pick for you or Google bot feels like
it has to pick for you. MARTIN SPLITT: And
it makes everything more complicated on
your side as well. RACHEL COSTELLO: Yeah. MARTIN SPLITT: Especially
if you're using things like search console, right? We are, we're gathering
data and showing you data based on the canonical. So if it starts flapping
between two URLs, then that's going to
look really weird. RACHEL COSTELLO: Mm-hmm. MARTIN SPLITT: So
anything else that you would say is unclear
about it or is there something that makes your
life really hard when it comes to kind of canonicalization? RACHEL COSTELLO: I
think it's figuring out that certain thresholds
you need to get to override Google's decision
on what is the preferred URL.
Because we can align all
of our signals on site. I saw that John Mueller on
the Ask Google Webmaster video about canonicalization, he
said that there's two aspects. You've got kind of
the on site signals, but you've also got what Google
thinks that the user would most like to have a look at. MARTIN SPLITT:
That's true, yeah. RACHEL COSTELLO: Yeah. MARTIN SPLITT: That depends on
a bunch of different things. So for instance, we might
canonicalize one language version over the other. If you were telling
us that all of them are canonical at the same
time and they have pretty much the same content, especially
if it's in the same language, just for different
countries, then we might show the version
to the searcher that the searcher is
in the country of. So if we have a DE
version and an AT version, of the German version
and an Austrian version, that are pretty much the same.
They use the same currency. That might have even the same
price, if you're unlucky. We might show different
URLs to searchers, depending on where
they are from. It makes more sense for
a customer in Austria to see the Austrian
version of the website rather than the German one,
even though the German one is the canonical. So that might be a little
confusing and misleading. RACHEL COSTELLO: Mm-hmm. MARTIN SPLITT: Any other
questions from your side? RACHEL COSTELLO: Yeah, so there
was one question I had in that. So if Google accepts the
canonical tag on a page, that it will ignore any
unique content on that page. But then that's interesting,
because surely, the pages have to be identical
in the first place. This is something I've heard. If there's any unique content
on the canonicalized page, it'll be ignored.
So how would that work? Would the canonical tag
not be accepted then, because they're slightly
different pages? MARTIN SPLITT: So that
depends on how different the unique content is. If you have mostly
the same content and then maybe have
like one sentence that is slightly
different, then we might still think that it's
pretty much the same thing. And then we would not see the
unique content necessarily if we think that it's just
a copy of another page that is canonical. If this page has the
canonical, then we would probably see the
unique content there as well, because it's the
page that we picked. However, if the content
is completely different or different enough
for the algorithms to decide that this
is not a duplication, then the canonical is pointless. RACHEL COSTELLO: Mm-hmm. MARTIN SPLITT: Unless there's
another page that happens, or another URL that happens to
point to the exact same page. Then it becomes
interesting again, because then we have
two different pointers to the same thing. And we get that
oftentimes that people are like linking [? to ?]
pages and accidentally have like some, I don't
know, some URL parameter that basically gets ignored or
doesn't actually matter, or there's like a slightly
difference in the way that the URL looks like.
Maybe you have like a slash
de, something something, and then like slash de,
something something question mark, cache equals
falls or something like that, that
doesn't really matter. Then we might canonicalize one
of these pages, and probably the one that does not
have parameters and stuff. That also is debatable. It might also happen that
we canonicalize something with parameters. But that way, you're
again making it harder for us to pick a canonical,
because if you're not saying like, oh,
this is specifically the canonical we want, then
it's back to guesswork. RACHEL COSTELLO: Mm-hmm. And I think that's the problem. People are just
trying to group pages topically maybe with
canonicalization, but that's not how it works.
MARTIN SPLITT: That's
not how it works, no. RACHEL COSTELLO: Thank
you for confirming that. MARTIN SPLITT: Canonical
tags and canonicalization is about reducing duplication. RACHEL COSTELLO: Yes. MARTIN SPLITT: That's
what it is for. RACHEL COSTELLO: Exactly MARTIN SPLITT: Awesome. Rachel, thank you so much
for being here and talking a little bit about
canonicalization with me. And I think that was useful.
And I hope you enjoyed it. Have a good time. Bye bye. [MUSIC PLAYING] Hey, everyone. I hope you liked the
previous episode. Next episode, me
and Glenn are going to discuss site moves, right? GLENN: Site moves, domain
name changes, URL migrations and more. MARTIN SPLITT: So stay
tuned and check it out..
